Rates of neonatal death and cerebral palsy associated with fetal growth restriction among very low birthweight infants. A temporal analysis
Dr A Spinillo, Department of Obstetrics and Gynaecology, IRCCS Policlinico S. Matteo, University of Pavia, P. le Golgi, 2. 27100 Pavia, Italy. Email firstname.lastname@example.org
Objective To assess whether changes over time in neonatal survival and infants’ neurodevelopmental outcome among very low birthweight (VLBW) infants was correlated with the obstetric aetiology of VLBW.
Design A cohort study of 773 VLBW infants.
Setting A University hospital in Northern Italy.
Population All the VLBW infants born over a 20-year period (1983–2002) at a single institution.
Methods Evaluation of neonatal mortality and neurodevelopmental outcome of the surviving infants at 2 years of corrected age. Logistic regression analysis was used to compare the improvements of neonatal outcome associated with obstetric risk factors over time.
Main outcome measures The risk reduction of neonatal death or cerebral palsy associated with each obstetric category responsible for VLBW over time.
Results The overall rates of neonatal mortality and cerebral palsy were 38.7% (43/111) and 17% (9/53) in the period 1983–87 and 13.7% (34/24) and 6.3% (13/205) in the period 1998–2002, respectively. The adjusted decrement per 5-year period was 33.1% (95% CI = 7.9–51.4) for neonatal death and 29.1% (95% CI = 25.3–32.7) for cerebral palsy, respectively. The adjusted rise in the rate of intact survival at 2 years of corrected age was 7.6% (95% CI = 3.1–12.3) per quinquennium. In logistic models with neonatal death or cerebral palsy as a combined outcome variable, and gestational age, corticosteroid use, surfactant use, and time of birth as explanatory variables, fetal growth restriction (P < 0.001) and pre-eclampsia (P= 0.011) interacted significantly with period of birth. The adjusted decrement in the rate of neonatal death or cerebral palsy as a combined variable was 27.5% per 5 years (95% CI = 13–39.6) in the overall population, 54.5% per 5 years (95% CI = 46.8–61.2) (P < 0.001 compared with overall population) among growth-restricted infants and 50.3% per 5 years (95% CI = 42.5–57.1) (P= 0.003 compared with overall population) in infants born to mothers with pre-eclampsia.
Conclusions Over a period of 20 years, the decrement in the rate of neonatal death or cerebral palsy was higher in growth-restricted fetuses than in other VLBW infants. This reduction was not obtained at the expense of an increased rate of neurodevelopmental impairments in surviving infants.
In the past two decades, progress in perinatal care has resulted in increased survival for very low birthweight (VLBW) infants (<1500 g).1 These advances have been attributed mainly to the regionalisation of perinatal care, the widespread antenatal use of corticosteroids for women at risk for preterm delivery, postnatal surfactant treatment for neonatal respiratory distress syndrome and improvements in the equipment for provision of neonatal intensive care.2–5 In obstetrics, although some progress has been made in the prevention of pre-eclampsia, little has changed in the pharmacologic treatment of preterm labour, prelabour rupture of membranes and hypertension complicating pregnancy.6 A wide variety of agents have been advocated for the suppression of uterine contractions, but ritodrine hydrochloride has remained the most widely used drug until recently.6 Similarly to almost 20 years ago, antibiotics recommended for the prevention and treatment of intrauterine infection include penicillin or erythromycin, and control of hypertension during pregnancy is mainly achieved using hydralazine, labetalol and nifedipine.6 On the other hand, there is good evidence that evaluation of fetal health by ultrasound and Doppler velocimetry in high-risk pregnancies can help in the timing of delivery, reducing perinatal mortality and morbidity.7,8 The causes of VLBW are very heterogeneous, including both spontaneous preterm birth and infants delivered electively for prelabour rupture of the membranes, pre-eclampsia or fetal growth restriction. The overall survival and neurodevelopmental outcome of VLBW infants is often reported according to gestational age and birthweight.1–4 However, the evaluation of the relative contributions of the obstetric factor associated with VLBW on survival and on the infant neurodevelopmental outcome might help us to identify pregnancies at increased risk and thus improve antenatal care. We analysed the relative impact of the various obstetric risk factors associated with VLBW on neonatal survival and infant neurodevelopmental outcome in an historical cohort of infants delivered during 20 years at a single institution.
The study used a database containing obstetric and infant information of an historical cohort of all VLBW infants delivered at the Department of Obstetrics and Gynaecology of the University of Pavia during the period 1983–2002. The study was approved by the review board of our institution and all the women gave their informed consent to the study. Maternal, social and clinical variables were collected at birth and entered into the database. Neonatal variables were collected at the discharge from hospital. Thirty-one out of the 804 infants (3.9%) with severe congenital malformation during the study period were excluded from the analysis. Gestational age was confirmed by an ultrasound scan obtained by the 18th week of pregnancy in 728 (94.2%) infants. In the 45 infants with uncertain gestational age and delivered in the first 5 years of the cohort, gestational age was assigned on the basis of neonatal examination. Fetal growth restriction was diagnosed when abdominal circumference measurements decreased to below the 10th percentile of our growth curve on two consecutive ultrasonographic evaluations taken 2 weeks apart. Our growth curve is employed by several years and was fitted after a longitudinal study of 657 normal pregnancies followed by biweekly ultrasonographic evaluations from 18 weeks of gestation to term. Pre-eclampsia was defined according to standard criteria.9 Nonreassuring antepartum and intrapartum fetal heart rate tracings were defined using the criteria described by MacDonald et al.10 The diagnosis of chorioamnionitis was clinical and was based on maternal temperature higher than 37.8°C plus one of the following features: 1. maternal tachycardia (>100 bpm), 2. baseline fetal heart rate greater than 160 bpm, 3. endocervical or high vaginal swab culture positive for a pathogenic organism, 4. maternal white blood cell count higher than 18 000/mm3, 5. foul smelling amniotic fluid. These criteria are similar to those recommended by Gibbs and Duff.11 Over the study period, ritodrine was the only tocolytic agent used, and the antihypertensive agents used in pre-eclampsia were methyldopa, hydralazine, and nifedipine.
Betamethasone or dexamethasone, 12 mg, two doses, 24 hours apart was the standard antenatal treatment for fetal lung immaturity until 1999; thereafter, betamethasone was the only steroid used. All the infants were transferred to our neonatal intensive care unit. Postnatal treatment with surfactant was routinely available by 1991.
Neurodevelopmental evaluation of the infants was carried out by a child neuropsychiatrist who was not involved in the intensive care of the infants. Examinations were carried out at discharge and after 3, 6, 12 and 24 months of corrected age. Neurological evaluation of the newborns was based on the methods of Amiel-Tison and Grenier.12 The Bayley scales of infant development were used to assess cognitive development (mental developmental index [MDI] at 12 and 24 months). Infants were grouped in three categories according to their final examination:
- 1Unimpaired: normal neurological status and Bayley MDI >84.
- 2Minor impairment: abnormalities of tone or reflexes but functionally normal or borderline (71–84) Bayley MDI.
- 3Moderate to severe impairment: neurological signs compatible with the diagnosis of cerebral palsy such as spastic diplegia or hemiplegia with a Bayley MDI ≥71 or spastic tetraplegia and/or severe mental retardation (Bayley MDI <71).
Of the 604 surviving infants (78.1%), 534 (88.4%) completed the neurodevelopmental follow-up programme. The 70 infants who did not complete the programme had been judged neurologically normal either at the discharge from hospital or at the intermediate visits. In addition, further information obtained by their family paediatricians confirmed that they had not developed later neurodevelopmental problems. However, to avoid any possible bias, these infants were excluded from the analysis of the association between antenatal factors and subsequent neurodevelopmental outcome.
Univariate analysis consisted of a chi-square test for association or linear trend for categorical variables and one-way analysis of variance for continuous variables.13 Unconditional logistic regression was used to evaluate the effect of obstetric variables on neonatal outcome adjusting for potential confounders, to test for linear trend and to compute adjusted rates of neonatal outcomes.14
Given their significant effect on neonatal outcome,1–6 gestational age, antenatal corticosteroid and postnatal surfactant administration were used as potential confounders in all logistic equations. Logistic models were also used to measure the relative impact of obstetric risk factors on neonatal outcome during the 20 years of observation. In order to study this relationship, we evaluated first-order interaction between period of birth and obstetric risk factors by including in logistic regression equations an interaction term (product coefficient term) between the putative risk factor (i.e. fetal growth restriction, preterm prelabour rupture of the membranes, etc.) and the period of birth. Obstetric risk factors were inserted as binary variables whereas year of birth was coded either as a discrete (annual) or binary (first and second decade) variable. The corresponding coefficients, including interaction terms, were further adjusted for the effect of the confounders.14 It has been suggested that for tests of interaction, significance should be set higher than the usual P value of 0.05.15 In the current study, an interaction was considered significant if its inclusion improved the model fit (P < 0.01) on the basis of likelihood ratio test.14 Since this was not a randomised study, we did not compute a preliminary sample size. However, given the prevalence of death or cerebral palsy in the overall population and among infants born after fetal growth restriction or to mothers with pre-eclampsia, sample sizes of 100 cases and 100 controls achieve a 93% power to detect a difference of 0.25 between the proportions of the measured event assuming four sequential tests on proportions.
The rates of the different neonatal outcomes tabulated by 5-year period are reported in Table 1. After adjustment for the effect of gestational age, antenatal corticosteroid and postnatal surfactant use, the drop in the rate per 5 years was 33.1% (95% CI = 7.9–51.4) for neonatal death and 29.1% (95% CI = 25.3–32.7) for cerebral palsy, respectively. Multivariate analysis of linear trend confirmed a significant trend in the reduction of both neonatal death (P for trend < 0.0001) and cerebral palsy (P for trend = 0.002). The rate of intact neonatal survival at 2 years of follow up increased from 49.1% (26/53) in the first 5-year period to 77.6% (159/205) in the last (multivariate P for trend < 0.0001). After adjustment for potential confounders, the relative rise in the rate of intact survival per quinquennium was 7.6% (95% CI = 3.1–12.3). When the analysis was restricted to infants of extremely low birthweight (ELBW) (<1000 g) (Table 1), the rate of death decreased from 54.1% (20/37) in the period 1983–87 to 33.7% (31/92) in the period 1998–2002. The adjusted decrement in the rate of neonatal death per 5 years was 16.3% (95% CI = 0.2–29.8). On the other hand, the adjusted increment in the rate of intact survival per quinquennium was 14.9% (95% CI = 1.7–29.2).
Table 1. Rates of neonatal death, CP, minor neurodevelopmental abnormalities and intact neurodevelopmental outcome among VLBW (<1500 g) and ELBW (<1000 g) infants born during the periods studied
|VLBW||n= 111||n= 205||n= 208||n= 249|| |
|Death; n, %, (95% CI)||43, 38.7, (30.4–47)||55, 26.8, (21.1–32.4)||37, 17.8, (13–23.4)||34, 13.6, (9.8–18.4)||<0.001|
|ELBW||n= 37||n= 99||n= 83||n= 92|| |
|Death; n, %, (95% CI)||20, 54.1, (38–69.5)||41, 41.4, (32–51.3)||27, 32.5, (23.1–43.1)||31, 33.7, (24.6–43.8)||0.04|
|VLBW||n= 53||n= 127||n= 149||n= 205|| |
|CP; n, %, (95% CI)||9, 17, (8.6–28.9)||19, 15, (9.5–22)||9, 6, (3–10.8)||13, 6.3, (3.6–10.3)||0.002|
|Minor; n, %, (95% CI)||18, 34, (22.2–47.4)||21, 16.5, (10.8–23.7)||35, 23.5, (17.2–30.8)||33, 16.1, (11.5–21.6)||0.09|
|Intact; n, %, (95% CI)||26, 49.1, (35.8–62.4)||87, 68.5, (60–76.1)||105, 70.5, (62.8–77.4)||159, 77.6, (71.5–82.9)||0.004|
|ELBW||n= 15||n= 52||n= 51||n= 59|| |
|CP; n, %, (95% CI)||2, 13.3, (2.3–37.5)||8, 15.4, (7.4–27.1)||6, 11.8, (4.9–22.9)||6, 10.2, (4.2–20)||0.13|
|Minor; n, %, (95% CI)||9, 60, (34.5–81.9)||8, 15.4, (7.4–27.1)||6, 11.8, (4.9–22.9)||11, 18.6, (10.2–30.1)||0.05|
|Intact; n, %, (95% CI)||4, 26.7, (9.1–52.5)||36, 69.2, (55.8–80.6)||39, 76.5, (63.4–86.6)||42, 71.2, (58.7–81.6)||0.026|
Table 2 reports the rates of neonatal outcomes according to the obstetric risk factor studied. After adjustment by logistic regression for potential confounders, spontaneous labour, vaginal birth and chorioamnionitis were associated with an increased risk of neonatal death or cerebral palsy measured as a combined outcome. Antenatal corticosteroids and elective delivery were protective factors.
Table 2. Rates of intact survival, minor neurodevelopmental abnormalities, CP and neonatal death among 703 VLBW (<1500 g) infants according to obstetric risk factors
|Gestational age (weeks)||30.2 (2.8)||29.6 (3.1)||29 (3.6)||27.4 (3.5)||<0.001|
|Birthweight (g)||1148 (281)||1131 (254)||1049 (270)||899 (289)||<0.001|
| ||n (%)||n (%)||n (%)||n (%)||OR (95% CI)**|
|FGR (n= 260)||153 (40.6)||44 (41.1)||11 (22)||52 (30.1)||1.37 (0.92–2.05)|
|Corticosteroids (n= 402)||229 (57)||71 (66.3)||31 (62)||71 (42)||0.6 (0.43–0.84)|
|Chorioamnionitis (n= 70)||33 (8.8)||12 (11.2)||9 (18)||16 (9.5)||1.71 (1.0–2.96)|
|Twins (n= 132)||83 (22)||8 (7.5)||8 (16)||33 (19.5)||1.02 (0.67–1.57)|
|Spontaneous labour (n= 275)||99 (26.3)||39 (36.4)||28 (56)||109 (64.5)||2.58 (1.8–3.72)|
|Pre-eclampsia (n= 178)||120 (31.8)||26 (24.3)||5 (10)||27 (16)||0.67 (0.43–1.06)|
|Elective delivery (n= 346)||227 (60.2)||65 (60.7)||13 (26)||41 (24.3)||0.39 (0.26–0.56)|
|Vaginal birth (n= 271)||102 (27)||39 (36.4)||27 (54)||103 (60.9)||2.18 (1.5–3.2)|
|PROM >72 hours (n= 70)|| ||11 (10.3)||7 (14)||17 (10)||1.02 (0.79–1.3)|
|Nonreassuring fetal heart rate (n= 131)||99 (26.3)||8 (7.5)||7 (14)||17 (10)||1.09 (0.7–1.69)|
|Third trimester haemorrhage (n= 74)||43 (11.4)||7 (6.5)||5 (10)||19 (11.2)||1.14 (0.67–1.93)|
To study the impact of obstetric risk factors on neonatal outcome in the different periods of birth, we studied first-order interactions between obstetric variables and year of birth adjusting for confounders. In logistic equations, time was modelled either as a discrete (year of birth) or binary (decade of birth) variable.
Gestational age was 29.3 ± 3.3 weeks in the decade 1983–92 and 29.4 ± 3.5 in the decade 1993–2002 (P= 0.7). The rates of pre-eclampsia (55/278 versus 123/407, χ2= 9.35, P= 0.002), elective delivery (86/278 versus 260/407 χ2= 71.7, P < 0.001) and antenatal corticosteroids (97/278 versus 305/425, χ2= 77.4, P < 0.001) were significantly lower in the first decade of birth compared with the second. On the other hand, vaginal birth was more common in the first period than in the second (160/278 versus 111/407, χ2= 52.4, P < 0.001). The rate of postnatal surfactant use in the past decade was 30.8% (131/425).
In logistic models, with neonatal death or cerebral palsy as a combined outcome and after adjustment for confounders, only fetal growth restriction and pre-eclampsia interacted significantly with period of birth. The interaction term between fetal growth restriction and period of birth was significant both when time was modelled as binary (decade) (P < 0.001) and discrete (annual) (P < .001) variable. When the analysis was restricted to ELBW infants, the interaction between fetal growth restriction and period of birth was still significant both when time was modelled as discrete (P= 0.003) and binary variable (P= 0.038). In particular, the risk of neonatal death or cerebral palsy associated with fetal growth restriction was significantly increased in the first decade but not in the second (Table 3). The adjusted decrement per 5 years in the rate of death or cerebral palsy was 27.5% (95% CI = 13–39.6) in the overall population and 54.5% (95% CI = 46.8–61.2) among growth-restricted infants (P < 0.001 compared with overall population) (Table 4). When neonatal outcome was modelled as intact survival versus others and after adjustment for confounders, the interaction between fetal growth restriction and period of birth modelled as discrete variables was still statistically significant (P= 0.043).
Table 3. Odds ratios of neonatal death or CP as a combined outcome in VLBW (<1500 g) infants born after FGR or maternal pre-eclampsia stratified by decade of birth
|Adequate growth||83 (45.6)||99 (54.4)|| ||73 (28)||188 (72)|| |
|FGR||43 (44.8)||53 (55.2)||1.86 (1.03–3.35)||20 (12.2)||144 (87.8)||0.98 (0.53–1.86)|
|Normotensive||107 (48)||116 (52)|| ||80 (26.5)||222 (73.5)|| |
|Pre-eclampsia||19 (34.5)||36 (65.5)||0.72 (0.38–1.36)||13 (10.6)||110 (89.4)||0.4 (0.22–0.76)|
Table 4. Rates of either neonatal death or CP among overall VLBW infants (<1500 g), VLBW infants with FGR and VLBW infants delivered after pre-eclampsia
|VLBW||n= 96||n= 182||n= 186||n= 239|| |
|Death or CP; n, %, (95% CI)||52, 54.2, (44.1–63.9)||74, 40.7, (33.7–47.9)||46, 24.7, (18.9–31.3)||47, 19.7, (15.0–25.10)||<0.001|
|FGR||n= 39||n= 57||n= 90||n= 74|| |
|Death or CP; n, %, (95% CI)||26, 66.7, (50.9–80)||17, 29.8, (19.1–42.6)||16, 17.8, (10.9–26.7)||4, 5.4, (1.7–12.5)||<0.001|
|Pre-eclampsia||n= 14||n= 41||n= 64||n= 59|| |
|Death or CP; n, %, (95% CI)||10, 71.4, (44.6–90.2)||9, 22, (11.3–36.5)||8, 12.5, (6–22.4)||5, 8.5, (3.2–17.8)||<0.001|
The risk of neonatal death or cerebral palsy associated with pre-eclampsia was significantly reduced in the second decade but not in the first (Table 3). In logistic regression analysis with neonatal death or cerebral palsy as the outcome, after adjustment for confounders, the interaction term between pre-eclampsia and period of birth was significant when time was modelled as a discrete variable (P= 0.011) but was of borderline significance when time was modelled as binary variable (P= 0.1). The adjusted decrement in the rate of neonatal death or cerebral palsy per 5 years among infants born to mothers with pre-eclampsia was 50.3% (95% CI = 42.5–57.1, P= 0.003 compared with overall population) (Table 4). Similarly to fetal growth restriction, the interaction between pre-eclampsia and period of birth modelled as discrete variable was statistically significant (P= 0.04) even when neonatal outcome was dichotomised as intact survival versus others.
Discussion and conclusions
The rates of neonatal death and cerebral palsy in this study are similar to those reported in other historical cohorts of VLBW from developed countries and in other population studies.1–4 Contrary to the data of Wilson-Costello et al.16 but similar to other authors,17,18 in our series, the increase in neonatal survival was not mirrored by an increase in the rates of severe neurodevelopmental impairment. The lack of a geographically based database of premature or VLBW infants is one of the main limitations of the present investigation. Although the results cannot be generalised as in population based studies, this study has the advantages of evaluating a uniform population of VLBW, detailing infant neurodevelopmental outcome.
During the period of the study, with the exception of an increased frequency of antenatal corticosteroid use, the pharmacologic treatment of pre-eclampsia, threatened preterm labour and preterm prelabour rupture of membranes at our institution has remained unchanged. As in other countries, postnatal surfactant was available by 1991. Since antenatal corticosteroid use and postnatal surfactant treatment significantly increase survival and reduce the complications of prematurity,4,5 we included these variables in the multivariate study of interactions. In logistic analysis, after adjustment for potential confounders, several interaction terms modified significantly the likelihood of several infant outcomes. During the period of the study, the risk of neonatal death or cerebral palsy associated with fetal growth restriction and pre-eclampsia decreased more than expected compared with the whole population of VLBW infants. During the same period, in these obstetric categories, the likelihood of intact infant survival at 2 years of follow up increased more than that of the other VLBW infants. This suggests that the increased survival was not obtained at the expense of an increased rate of neurodevelopmental impairment of the infants. It is important to point out that logistic models evaluate multiplicative interactions between variables, or more simply the departure from an additive relationship. It is therefore possible that neonatal outcome was improved also for other obstetric categories, although this improvement was less significant than that recorded for fetal growth restriction and pre-eclampsia. In geographically defined population studies, fetal growth restriction among VLBW infants is associated with a consistent increase in the risk of neonatal death and complications of prematurity.19 The increase in the risk of a poor outcome among growth-restricted VLBW infants is probably a direct consequence of fetal hypoxaemia, nutrient restriction and altered endocrine environment associated with uteroplacental ischaemia.5,19 The relationship of pre-eclampsia with short-term and long-term infant outcome is uncertain.20 Although in some studies, pre-eclampsia is associated with a reduced risk of periventricular leucomalacia and subsequent cerebral palsy,21,22 other large series have not confirmed this protective effect.23 Nonetheless, the improved outcome among infants delivered after pre-eclampsia or fetal growth restriction is numerically important because these two categories comprise more than 40% of VLBW infants.
The reason for an improved outcome among growth-restricted VLBW infants and associated with pre-eclampsia could be related to an improvement in the timing of delivery associated with the widespread use of ultrasound evaluation of fetal health, Doppler velocimetry and computerised fetal heart rate tracing. Data from randomised trials suggest that the use of umbilical artery Doppler studies in high-risk pregnancies, especially those complicated by hypertension or presumed impaired fetal growth, is associated with a trend to a reduction in perinatal deaths, fewer inductions of labour and fewer admissions to hospital.6,7 The availability and use of umbilical artery Doppler studies for assessment of the fetal–placental circulation in pregnant women with suspected placental insufficiency has been now included in some standard obstetric guidelines.7,8
In conclusion, over a period of 20 years, the decrement in the rate of infant death or cerebral palsy was higher in growth-restricted fetuses than in other VLBW infants. This reduction was not obtained at the expense of an increased rate of neurodevelopmental impairments in surviving infants.